1

I have embedded software with the following version-numbering :

<Product Number>.<Major>.<SVN commit number>

  1. Product Number is arbitrary set up and corresponds to each soft
  2. Major is arbitrary set up and it is incremented with major evolution of the soft
  3. SVN commit number is incremented with each commit

The problem is that we swapped from svn to git and we lost the SVN commit number which is "replaced" by SHA.

Thus, we could replace <SVN commit number> by <git commit SHA>. However, we can't do that because we are constrained to 2 bytes for this last versionning-number, and <git commit SHA> is often compressed to 3 bytes.

Is there another way to have a version-numbering taking into account commits?

Clément
  • 1,128
  • 7
  • 21

2 Answers2

1

If you accept the possibility to have an hash composed of just 2 bytes you can use the options --abbrev=<n> of the git describe command where n is the number of bytes you want for your git hash. In your case it would be:

git describe --always --abbrev=2

Remember that there is the possibility that with just 2 bytes your hash is no more unique, even if it is a remote chance given that it is also related with you version number. The --always option is needed to show only the hash.

LeonFibuck
  • 336
  • 2
  • 12
1

You can use git rev-list --first-parent --count @ to get a count of first-parent commits in your checked-out history, sub in a release tag for @ to get that release's count.

If you can work out from the product and major numbers which tip you were on, say you've got 143.3.9234 and you maintain product 143, major 3 as the wizzo-v3 release branch, you can git rev-list --first-parent --reverse wizzo-v3 | awk NR==9234.

And of course there's always the option of using that number in your release tag name.

jthill
  • 55,082
  • 5
  • 77
  • 137